Developing state policies on AI and autonomous vehicles is crucial to effective federal regulation.
In a recent essay in The Regulatory Review, Richard J. Pierce, Jr., urged the Trump Administration to issue rules that preempt state laws and regulations. He explained this was necessary to avoid what he calls “a chaotic combination of inconsistent rules” that may “severely impair efforts” to develop artificial intelligence (AI) and autonomous vehicles. This path could be a “one in, fifty out” deregulatory agenda, Pierce argues, akin to the Administration’s executive order requiring federal agencies to eliminate 10 rules for every one it issues. I write to offer a defense of regulatory federalism.
First, it is important to recognize the threat that aggressive preemption by the federal government poses to effective regulation. In the lead-up to the 2008 financial crisis, state regulators sought to police the unfair and abusive lending practices they saw in their states, particularly in the subprime mortgage market. National banks were major funders of subprime mortgage issuers. The Office of the Comptroller of the Currency (OCC), however, issued a rule preempting the states from enforcing rules against nationally chartered banks, meaning the states could not enforce consumer protection laws against them. Former Illinois Attorney General Lisa Madigan later told the Financial Crisis Inquiry Commission that the OCC was “particularly zealous in its efforts to thwart state authority over national lenders, and lax in its efforts to protect consumers from the coming crisis.”
Without state-level enforcers, predation and fraud blossomed, and the economic consequences were devastating. When the U.S. Congress passed the Dodd-Frank Wall Street Reform and Consumer Protection Act, it limited the OCC’s authority to preempt state consumer financial protection laws by providing a specific preemption standard and declared decisively that federal banking law “does not occupy the field in any area of State law.” Congress further provided that, in evaluating whether state consumer financial protection laws are preempted, a state law will not be deemed inconsistent with federal law so long as the protection it provides to consumers “is greater than the protection provided” by the Dodd-Frank Act.
Second, the extent to which differences in state regulation lead to chaos and inconsistency is probably overstated. It is reasonable to assume that jurisdictional differences can lead to challenges and added compliance costs. In practice, however, states use several strategies to harmonize their regulatory programs. For example, the Conference of State Bank Supervisors established a joint state examination system that allows for coordination, data sharing, and standardized processes for the examination of state-regulated financial institutions. This system allows regulated entities to engage with state financial regulators in a more orderly, coordinated fashion and allows state regulators to enforce their own laws more efficiently.
State attorneys general frequently file joint litigation to address common concerns, sometimes alongside or in coordination with federal regulators such as the Federal Trade Commission (FTC). In some areas, such as consumer protection, state laws are modeled after federal laws. The FTC Act creates a body of law that broadly applies across states, even if the particulars vary. There are also organizations, such as the Uniform Law Commission, that provide model laws for states.
Finally, state regulatory programs provide valuable information to Congress and federal agencies about what federal standards should be, if any. In its report accompanying the Dodd-Frank Act, the U.S. Senate Committee on Banking, Housing, and Urban Affairs observed that preempting state-level consumer protection laws would deprive the federal government of “an important source of information and reason to adjust standards over time.” When a state legislature enacts a law, it sends a clear signal to members of Congress from that state about what their constituents want.
This is true across areas of regulation, including artificial intelligence. For example, U.S. Senator Marsha Blackburn (R-Tenn.) retreated from a compromise amendment to the AI moratorium in the Senate’s version of the recently passed law due in part to concerns about protections against the unauthorized use of artists’ voice and likeness, which Tennessee had provided in its state-level Ensuring Likeness, Voice, and Image Security Act. State-level regulation also provides opportunities for regulated industries, interest groups, and the broader public to voice concerns, which may eventually inform federal legislation.
Not all preemption is the same. Generally speaking, under the Supremacy Clause of the U.S. Constitution, there are two kinds of preemption: conflict preemption, where a state law is preempted when it is impossible to comply with both state and federal law, and field preemption, where the federal government has chosen to wholly displace state law, even if it were theoretically possible to comply with both. Immigration law is an example of an area where the federal government is understood to have completely occupied the field, leaving no place for the states to enact their own laws. In some cases, such as in the Dodd-Frank Act, Congress explicitly channels state regulation in particular ways by delineating when and how states may regulate on top of a federal scheme.
It is not entirely clear whether Pierce envisions the Trump Administration pursuing field or conflict preemption with respect to artificial intelligence and autonomous vehicles. On the one hand, urging the administration to preempt merely “inconsistent” state law suggests a more modest conflict preemption approach. His call for “comprehensive” regulations, however, suggests a field preemption strategy. At this stage, asking the federal government to completely occupy the field of regulation for these technologies is most likely an insurmountable task. Artificial intelligence has a vast array of potential applications across numerous realms of existing law, including finance, labor, consumer protection, and law enforcement, just to name a few. It seems difficult to imagine devising or adapting regulations to all potential applications in anything but an iterative way.
In sum, hindering regulatory development by preventing states from developing their own laws on AI and autonomous vehicles risks depriving policymakers of important information and the public of opportunities to shape the future of these technologies. As the 2008 financial crisis demonstrated rather tragically, zealous preemption may also have unforeseen consequences.
Pierce qualifies his argument by saying the Trump administration should preempt state regulation through the implementation of “reasonable” rules to govern AI and autonomous vehicles. State-level policy development and implementation are important processes for understanding what “reasonable” rules are.
